DT
/"i;r-'T3:.'T
:aBOOM
'L
U
.z:2Ch,U'
3
W
36-412
#t7
C-T ot of 2ChnlQog
THE USE OF FREQUENCY MODULATION
FOR TELEVISION TRANSMISSION
E. E. MANNA
G. M. RODGERS
R. D. STUART
L. B. ARGUIMBAU
TECHNICAL REPORT NO. 259
JULY 6, 1953
RESEARCH LABORATORY OF ELECTRONICS
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
CAMBRIDGE, MASSACHUSETTS
r
The Research Laboratory of Electronics is an interdepartmental laboratory of the Department of Electrical Engineering
and the Department of Physics.
The research reported in this document was made possible
through support extended the Massachusetts Institute of Technology, Research Laboratory of Electronics, jointly by the Army
Signal Corps, the Navy Department (Office of Naval Research),
and the Air Force (Air Materiel Command), under Signal Corps
Contract No. DA36-039 sc-100, Project No. 8-102B-0; Department of the Army Project No. 3-99-10-022.
t
MASSACHUSETTS
INSTITUTE
OF
TECHNOLOGY
RESEARCH LABORATORY OF ELECTRONICS
July 6, 1953
Technical Report No. 259
THE USE OF FREQUENCY MODULATION
FOR TELEVISION TRANSMISSION
Manna
R. D. Stuart
E.
L. B. Arguimbau
G. M. Rodgers
E.
Abstract
We have investigated the application of frequency modulation to television transmission. We have tested the possibility of improving the picture quality by using a receiver
capable of taking full advantage of the capture effect.
sarily provide a distortionless signal.
This procedure does not neces-
A ghost, characteristic of frequency modulation,
is obtained under multipath propagation conditions.
It is quite different from the ghost
produced by amplitude modulation under similar conditions.
By the use of de-emphasis,
the effects of the frequency-modulation ghost can be reduced but not removed. A wider
bandwidth system can give a better reduction of the interference. The question of
whether the ghost obtained with frequency modulation is less objectionable than that
obtained with amplitude modulation is a subjective one and is difficult to answer.
*
_
_
THE USE OF FREQUENCY MODULATION FOR
TELEVISION TRANSMISSION
1.
Introduction
Previous work (1, 2, 3) has shown that the effects of interference caused by multi-
path propagation may be considerably reduced by the use of frequency modulation.
By
designing the receiver to take full advantage of the capture effect it is possible to obtain
distortionless operation in the presence of high-level, co-channel interference.
A theo-
retical investigation of this effect was carried out some years ago, and experiments on
the application to sound transmission were made (1,
2.
2, 3).
Theory of the Method
Consider the case of two-path transmission.
The time variation of instantaneous
frequency of the signals arriving at the receiver over the two different paths will be the
same, but one will be delayed relative to the other by an amount depending upon the difference in length of the two paths (see Fig. 1).
Over a short interval of time (such as that between t 1 and t 2 ) we can consider that
the frequency of the one signal has a constant value, p; the other, a constant value,
p + r.
Suppose that the amplitude of the signal of frequency p is unity and that the am-
plitude of the other signal is a (a is less than unity).
The resultant signal is then given
by the vector sum of these two components (see Fig. 2).
The interfering signal will
rotate with a frequency r relative to the wanted signal and will cause the amplitude
of the resultant to vary.
It will also cause the instantaneous speed of rotation of the
resultant to differ from that of the wanted signal, but since a is less than unity it is
easily seen that the total number of rotations round the origin made by the resultant is
the same as that made by the wanted signal.
In practice p is always large compared
with r,
and so the average frequency of the resultant is the same as that of the wanted
signal.
This is the "capture effect."
If the amplitude of the resultant vector is A, then we can write (see Fig. 2)
Aej
= 1 + aejrt
0 = Im [log (1 + aeJrt)]
The instantaneous frequency of the resultant is obtained by differentiating the phase, 0,
with respect to time.
do= Im [d log (
=ar
+ ae j r t )
+(a + cos rt)
1 + 2a cos rt + a
as a-l,
d/dt
- r/2, [cos rt -1].
When cos rt = -1, we have d/dt = (-ar)/(l-a).
-1-
__I__
TIME
Fig. 1
Variation of angular frequency
of direct and delayed signals.
Fig. 2
Vector diagram for twosignal interference.
A
II I
0)1
p+r
p+
<l
I
TIME
Fig. 3
Variation of instantaneous frequency
for two-signal interference.
a = 0.8 (solid line)
a = 1. 25 (dotted line)
-2-
I I
I
I I
Thus over the major part of the beat frequency cycle the frequency of the resultant
is approximately the mean frequency of the two components.
At the instant when these
components are in antiphase there is a large shift in frequency in the direction of the
stronger signal, which makes the average frequency of the resultant the same as that of
the stronger signal. A plot of instantaneous frequency against time is shown in Fig. 3.
The maximum difference between the instantaneous frequency and the frequency of the
stronger signal is given by (-ar)/(1-a) and is largest when r is largest. The largest
possible value of r occurs when the two signals are at opposite extremes of the range
of deviation. If the weaker and stronger signals are interchanged, the spikes are in the
opposite direction (shown dotted in Fig. 3).
The total range covered by the spikes is
therefore [(l+a)/(1-a)] rm where rm is the frequency deviation. When a = 0.7, this
range is nearly 6 times the deviation; when a = 0.95, it is 39 times the deviation.
To preserve the average frequency, it is necessary to preserve the spikes; therefore
the frequency detector must be linear over this wide range.
Further, for large values
of a,
the frequency changes very rapidly with time during the period of the spike, and
the detector must be fast-acting to prevent diagonal clipping. When the two components
are in antiphase, not only is there a rapid variation in instantaneous frequency, but there
is also a sharp dip in amplitude.
Thus the limiters must be able to remove this downward amplitude modulation, which also means that they, too, must be fast-acting. The
intermediate-frequency amplifiers must be flat over the deviation range to insure that
the stronger signal at the input always remains the stronger signal at the output, irrespective of its own frequency or that of the interfering signal.
always be some ripple in the passband.
In practice, there will
The worst situation will arise when the wanted
signal falls at a minimum of the characteristic and the interference falls at a maximum.
The ratio of unwanted to wanted signals at the output is then higher than that at the input
by a factor given by the ratio of maximum to minimum amplification in the passband.
Thus, if the percentage ripple in the passband is 100E, and the limiter and detector are
designed for a capture of a , then the over-all capture ratio of the receiver is given by
a = (l-0)al. It should be noted that the intermediate-frequency bandwidth is not determined by the same considerations which determine the total frequency range covered by
the spikes.
(In fact, it could not be, since then the receiver would have no selectivity.)
The signal in the intermediate frequency consists of two spectral components, one at
p and one at p + r.
These are obviously always within the range of deviation, and this
range, consequently, determines the required intermediate-frequency bandwidth. This
signal, however,
contains large spikes in its instantaneous frequency and also large
amplitude modulation. It is the removal of the amplitude modulation by the limiters
which gives rise to the higher-order spectral components and implies the need for
a broad bandwidth in all subsequent stages (i.e.,
in the limiters and the detector).
Although it is possible by these means to obtain from the receiver an output signal
whose average frequency is the same as that of the wanted signal, it does not necessarily
follow that this output will be distortionless. This depends upon the amount of filtering
-3-
it is possible to employ for the removal of the spikes.
We shall return to this point
later in connection with pre-emphasis.
3. Method of Experimenting
It was found convenient to make the experimental tests with a facsimile system
rather than a standard television system.
The particular facsimile system that we used
operates without synchronizing signals during the transmission of the picture.
means that attention is concentrated on the picture quality attainable.
This
The time taken
to scan one line in this system is about 0.6 sec; in television it is approximately
60
sec.
The scale factor in time is therefore about 1 : 104.
A block schematic diagram of the equipment is shown in Fig. 4.
The signal from
the facsimile transmitter is obtained as an amplitude-modulated carrier at a frequency
of 1800 cps.
This signal is detected and the video waveform used to frequency-modulate
a 14 kc/sec carrier, which can then be passed through a filter to limit the frequency
band used. To simulate multipath conditions, an acoustic delay line is used to give a
delay of some 25 msec. The two signals, direct and delayed, are passed through separate limiters and then added together in the desired ratio.
The resulting signal is then
passed through a limiter and a frequency detector and the video output fed, as amplitude modulation of a carrier of 1800 cps, to the facsimile receiver.
Although the experimental work has been done with a facsimile system, the figures
quoted in this report will be those that correspond to the television system.
A value of 140 Mc/sec was chosen as suitable for the carrier. The bandwidth
assigned for television channels in practice is 3 Mc/sec for single sideband operation.
Double sideband operation was employed here and a bandwidth of 5 Mc/sec was used.
Interfering signals up to 70 percent of the wanted signal were investigated; hence the
bandwidth required in the detector is about 28.5 Mc/sec.
In order to assess the
advantages of a wider transmission bandwidth, the detector is linear over a range of
60 Mc/sec. This allows for an interference of 70 percent and a maximum bandwidth of
about twice the present television standard.
A ratio detector was used to insure rapid-
ity of action and to take advantage of the limiting properties of this type of detector. It
is preceded by a three-stage crystal limiter. In this equipment no i-f amplifier is used.
The effect of multipath interference in television, when using amplitude modulation,
is to produce a ghost picture in addition to the main image.
A typical value for the
spacing of the two images on a 12-inch screen is approximately 0.5 inch, which corresponds to a delay of 2.5 ,psec.
In these experiments we used the whole of the available bandwidth for deviation.
Since this will not allow the transmission of the higher-order sidebands, intolerable
distortion of the picture might, at first sight, be expected. However, a picture waveform is composed mainly of fairly rapid transitions from one level of grey to another.
The most stringent case to be dealt with consists of a transition from black to white
with the maximum rise time of which the system is capable.
-4-
I
Tests were made with a
Fig. 4
Block schematic diagram of equipment.
lOW
87T
4w
2w
0
WHIT
z
_
fw
LL
fb BLA
A
' 'I
I,
' I
l
(d)
||
Fig. 5
|
The formation of frequencymodulation ghosts.
|
three-element Butterworth filter with 3-db points at 140 + 2.5 Mc/sec. The test signal
was frequency-modulated with a trapezoidal wave of 0.08 CFsec rise time. The output
of the filter was then detected and the rise time of the output waveform observed. When
the frequency swing was 5 Mc/sec the rise time increased to 0.2 Flsec. (A slight overshoot was also observed.) Thus in the worst case the rise time is increased by a factor
of only 2.5. Pictures were taken with and without this filter in the experimental system
and very little over-all difference in picture quality is noticeable.
-5-
-__
------
_ I_
___________
An important difference between the system shown in Fig. 4 and an actual transmitterreceiver system might be mentioned.
In section II we noted that the intermediate-
frequency stages of the receiver must have a flat frequency response to obtain effective
capture.
This, in fact, applies equally to all of the stages which are previous to the
intermediate-frequency stage, including the amplitude-frequency characteristics of the
transmitted signals.
However,
in the arrangement shown in Fig. 4, the two signals
are received separately and are passed through separate limiters.
two limiters are then added in the desired ratio.
The outputs of the
Thus the frequency characteristic of
the transmitting system produced by the filter in the transmitter, or by the characteristic of the delay network, cannot upset the capture properties.
But since the higher-
order sidebands removed by the filter cannot be recovered by limiting, the frequency
characteristic of the transmitting system can give rise to distortion.
4.
Results
We shall first discuss the effect of the spike interference on the picture.
Consider
a simple black-to-white transition, as shown in Fig. 5a, and suppose that we have twopath transmission.
Suppose,
also,
that the direct signal is
the larger.
Before the
transition takes place, the frequency of the signals arriving by both paths is that corresponding to black,
fb; therefore, the frequency of the resultant is fb (Fig. 5c).
For a
period equal to the time delay of the weaker signal on the stronger, the frequency of the
stronger signal is that corresponding to white, fw, whereas that of the weaker signal is
still fb.
Throughout most of this period,
then,
the resultant signal has a
)
1/2 (fb + fw which corresponds to some shade of grey.
However,
frequency
since the average
frequency of the resultant must be fw, there will be frequency spikes in the direction
corresponding to white.
These spikes will exceed the white level but since a
corresponding to "whiter than white" will be registered as white by the system,
resulting pattern will consist of white spots on a grey background.
(See Fig.
signal
the
5d.)
(Note that in Fig. 5, the abscissa can be thought of as either time or distance across
the picture; the two are related by the speed of scanning, which is constant.)
Now suppose that the time delay of the interfering signal is
period of "black" transmission,
T.
Then during a
the phase difference between the two signals is
~b = 2'rTfb + 2rk (where k is an integer).
When the direct path changes frequency,
the phase difference between the two signals increases linearly at a rate given by
2w (fw - fb )
as shown in Fig. 5b.
Thus,
if the instant at which the direct path changes
corresponds to t = 0, the phase difference at time t later is given by
2rTfb + 2rk + 2(fW - fb)t
A spike is formed when this expression has a value (2n - 1)1T,
(2n - 1)r = 2rTfb+ 2rrk+ 2(fw - fb)t
-6-
__
___
_
_
__
where n is an integer
or
12(f-b
= 2(fw 1 f)
[{2(n-k)-
[2(n-k)-
1}
1
-
-
22Tfb]
fb
The first spike is formed when
1 - 2Tf b
t -2(f
which is dependant only upon
T,
- fb
)
fw, and fb' Therefore the distance of the spot from
the transition is the same for all transitions between two given frequencies.
And on any
line scan, the white spots will fall adjacent to the ones formed on the previous line and
the resulting pattern will consist of white stripes on a grey background, as shown in
Fig. 5d.
Furthermore, the pattern formed on any particular frame will superimpose
exactly on that formed in the previous frame.
From section II we know that the spike recurrence rate is equal to the frequency
difference between the two signals.
When we use a 5 Mc/sec deviation, the spacing of
the stripes following a black-to-white transition, when viewed on a 12-inch screen, will
be approximately 0.04 inch, which will be quite noticeable.
A transition of lower con-
trast will give wider spacing and will lead to a more objectionable effect, although the
magnitude of the spike (which is also proportional to the frequency difference) is smaller, until the spike becomes small enough to make the variation in intensity negligible.
The problem of removing this effect by filtering is difficult.
low-pass filter would be to reduce the height of the sharp spike.
The major effect of a
This reduction will not
make any difference to the picture, since it will still be registered as white.
The filter
will also cause only a relatively small shift in the grey level of the background.
A more
promising method of removing the effect would be to increase the spike recurrence rate
above the resolving power of the system.
In fact,
if this resolving power is just
sufficient to resolve the highest frequency that we desire to transmit, this solution
automatically demands a bandwidth assignment greater than that determined from
the frequency components in the modulation waveform.
(This is the situation existing
in frequency-modulation sound broadcasting where the highest audiofrequency transmitted is about 15 kc/sec, and the bandwidth assignment is + 75 kc/sec.) If we retain
the original 5 Mc/sec bandwidth, the spike frequency will always be in the same region
as the picture information.
It would appear, therefore,
that the spikes cannot be
removed without removing some of the picture information, with consequent deterioration of the quality of the final picture.
However,
the spikes are formed at the
receiver; they are not present in the original modulation waveform at the transmitter.
Thus it is possible to design a filter for reducing the effect of the spikes
without affecting picture quality, if the inverse filter is used at the transmitter to
pre-correct for its effect.
We are therefore led to a consideration of pre-emphasis,
-7-
_
_____1_1
commonly used in frequency-modulation sound transmission.
Consider, again, the case of a rapid transition from black to white.
waveform will approximate a ramp function.
The resulting
If this is applied to a pre-emphasis cir-
cuit, which is essentially a differentiating circuit, an overshoot is produced.
If the nor-
mal black-to-white transition shifts the transmitter frequency from one end of the band
to the other, the pre-emphasized transition would, because of the overshoot, take the frequency outside the assigned bandwidth.
To prevent this from happening, the frequency
difference corresponding to a black-to-white transition would have to be reduced.
This,
however, would reduce the repetition frequency of the spikes, making them more difficult to remove. An investigation into the problems of video pre-emphasis has been
made previously (4).
The solution adopted here was to pass the pre-emphasized wave
through a clipping circuit that removes all overshoots that would take the transmitter
outside the prescribed range.
Since a standard de-emphasis circuit is used, some dete-
rioration of picture quality results.
The characteristic of the pre-emphasis circuit is
given approximately by[ + (f/fp)2] 1/2
By using a ramp function with a rise time of
0.08 rLsec this circuit gives an overshoot of about 100 percent, with fp adjusted to 800
kc/sec.
From earlier numerical calculations (using the uniform stretch function (5)
to represent the transition), we can estimate that removing the whole of this overshoot
and then using standard de-emphasis will increase the rise time by a factor of about 8.
Hence, there may be some blurring of high-contrast transitions.
case that can occur.
But this is the worst
The rise time of a transition less than full black-to-white will not
suffer so much deterioration,
since only part of the overshoot is removed.
contrast is less than a certain level, none of the overshoot is
removed.)
(If the
However,
the blurring of high-contrast transitions will always be present, even under single-path
transmission conditions. The characteristic of the de-emphasis circuit is given by
[1 + (f/fp)2]- 1/2. The lower the frequency fp, the more effective the de-emphasis circuit becomes in reducing the spikes; but the overshoot produced by the corresponding
pre-emphasis circuit will increase.
This overshoot must be removed to prevent the
given bandwidth being exceeded; and the larger that overshoot is,
rioration of picture quality caused by its removal.
the greater the dete-
Thus the retention of picture quality
and the reduction of the spikes are conflicting requirements.
5.
Representative Pictures
An original of the picture used for these tests is shown in Fig. 6.
The facsimile
system operates with a picture 8.5 inches by 7 inches which is scanned along its longer
dimension and uses approximately 600 lines.
Thus the resolving power of the facsimile
system is somewhat greater than that of the television system. A television screen is
normally viewed at a distance that makes the line structure unnoticeable. To get a proper comparison, the pictures reproduced here should be viewed at a distance of approximately 5 feet.
The first group of pictures (Fig. 7) was taken using amplitude modulation for
-8-
I
I
_
comparison.
The group of pictures shown in Fig. 8 illustrates frequency-modulation
ghosts produced by interfering signals of different strengths.
The whole of the 5 Mc/sec
bandwidth was used for deviation; a pre-emphasis frequency of 800 kc/sec was used; and
any overshoots that would take the transmitted frequency outside the assigned bandwidth
were removed.
In a typical practical case the interfering signal will be about half the
wanted signal; Fig. 9 shows the results of experiments made with this level of interference to determine whether there is any de-emphasis frequency that will give a satisfactory reduction of the spikes without intolerable deterioration of picture quality.
With the
lowest de-emphasis frequency used (200 kc/sec) the ghosts are still objectionable and
the picture quality is already unacceptable.
It would appear that with a 5 Mc/sec band-
width no satisfactory compromise can be made to reduce the effects of the ghosts caused
by an interference level of 50 percent without impairing the quality of the picture too
much.
It may be noted that with the present system used for television broadcasting no
effort is made to reduce the effects of ghosts at all.
By the same standards this system
would also be regarded as unsatisfactory.
The results of using wider bandwidths are shown in Fig. 10.
Here the ghosts
occurring at high-contrast transitions are removed almost completely without objectionable deterioration of picture quality, but the ghosts occurring at lower contrast transitions are still in evidence.
The de-emphasis circuit cannot effectively remove these
because the spike frequency is lower.
6.
Conclusions
Under conditions of multipath transmission, it is possible for the stronger signal to
capture in a frequency-modulation system even in the presence of a strong interfering
signal.
This does not necessarily result in an undistorted signal.
is present and its repetition rate depends upon the frequency swing.
mission the spikes produce a characteristic ghost.
Spike interference
In television trans-
If the assigned bandwidth is the same
as the highest components of the video waveform that is to be transmitted, the spike frequency always falls in a region of the spectrum where there is picture information, and
effective filtering is difficult.
not completely solve it.
The use of pre-emphasis simplifies this problem but does
If wider bandwidths are employed, then a more effective reduc-
tion of the interference is possible.
-9-
_lli__
111·_11_1_--
11--sL-11_1___1---LL--·--·.lll-_l
1··11·----(11--1I---^-·--·-XC
1I
I-
---
References
1.
L. B. Arguimbau, J. Granlund: Transatlantic FM Transmission, National Electronics Conference, Chicago, Nov. 1947
2.
L. B.
1949
3.
J. Granlund: Interference in Frequency-Modulation Reception, Technical Report
No. 42, Research Laboratory of Electronics, M.I.T. 1949
4.
W. L. Hatton: Pre-Emphasizing Video Signals, Technical Report No. 202, Research
Laboratory of Electronics, M.I.T. 1951
5.
H. E. Kallman, R. E.
33, 169, 1945
Arguimbau, J.
Granlund:
Sky-Wave FM Receiver, Electronics
Spencer, C.
P.
Singer:
-10-
22,
101,
Transient Response, Proc. I.R.E.
CAUTION
Figures 6-10 are meaningless as a gauge of the system
unless they are viewed from a distance of at least 5
It is very important to use this viewing distance.
feet.
This awkward procedure has been adopted to reduce
the effects of the printing process.
-11-
ill
I
I
11
_1_-111___
o
"-4
E/I
L
Q
u
.
k
)iU
*H
s4-
·
Er
*
-4 b.0
ri)
r.
0
U
bl
r. -4cd
a)
04
I
U
U
U
vIa
:1·:·1""
I
s
'``: d
·
i
_'
- 12-
0
.4
-.o
0
0
"0
LO
,-
C)
,.44
"-4
.-
0
o
I
.
o
o
C)
0c)
U2
-13-
___
_______11_1___1___
·1·^
I-1-_-
II_
I
o
o-I
cd
0
"0
.-
a
ct
c4
Cd
~~~~~~~
lsa ~ X~lq
.I
F
z
u)
.'
o0.)
4-.
II
0)
.-
.,'-t
-1
.,-
¼;
.
,W
I
i
-14-
_
o
-4
a.
0
0
0
.,
ab
a,-
.4
3
(D
:
a-
0
u]
cH
-15-
___
I
-
,Q
a)
ID
i b.0.
C-)
O
,
-41
W
,l
01)
OQ
I
Cd
"41
-'-4
0
.H
a)
r.~
.
o
04
b.0
E-H
-16-
'0
a
C-)
U
*4-4
o
C.
bo
oi)
c)
oa
lr
b.0
b.j
Q)
r·H
H
-17-
____11^11__1_
1·
--
·---
___I
c)
o
0
4
0+-'
-U)
LO
To U
1
0 ,.
u
EII
Cd)
bcj
ofHo'r
o
-18-
-4
-
0
Cd
a)
4-
a)
4
- b.-
a)
C
4 4
Ln
o
U
c,,
a
0
..o4
a)
o
U)
E--
-19-
lil_XII
------------
~
_II_--
s
-----
- --
I
C)
a4
V)
t1
Cd
c4
a)
a)
c
-4
4
aa
a
) LC
-o
0
c
4
co
a;
HrOe
0,
W,
4-
Q)
Q)
-20-
*I
a)
c)
a4
a)
-CI
4
a0
E
-
a)
a)
C.4la) Lf
U)
,
0.
a)
C)D
ocY
o
in
v~4.H
, ,,
C-
cr0
a
a
.Fi
B
.,
43
41
Ev
-21-
C)
(n
0.)
0L)
Ic
U4
·
4-
a) n
b0
r.
otn
4- 4
0
U
c
-4
4
0)
4-.
0
EH
-22-
-*
o
)
(1
-,-
U2
a,
US)
oI
_
U
II
a
C-)
U~
)
-4
Cd
o 1a
l
oU
E0
b0
0
F =
*B
l
o
11
,,
0,
F-
-23-
-
U
10
r
.Q
a)
o:
~
-
b
0o
(12
LO
a)
m U
12
'0
4
a)
o)
do
a)
-4)
a)
.Ov
II
-24-
___
. ,-I
o
u
Oa)
o
a)
[la
44
a)c
Q
a.
a)
a)
)
Li
U
E-4
-4
1%
-
-
-,
Ca
r
E
UO
O.*
41
zW ~~PdI O~~~r
pi~~~~~~,E~~~~0~0
0p.
u3
1
x
P-f
(::I
a.)Q
a
-~~~4Lfl
Cd~~~~a
~~~~~~~~si
C)
b0
O
0
o
-I
-4
0
a)
-25-
X ____I
I_
_____
a
·
C
.".
-
U
a)
O
s,
°
C
PC
S
0
a,
_.
:
VJ
A
'
.q
>
I
¢
'I
--
_